lkml.org 
[lkml]   [2010]   [Jul]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 5/5] Package Level Thermal Control and Power Limit Notification: pkgtemp doc
> > +This driver permits reading package level temperature sensor embedded inside
> > +Intel CPU package. The sensors can be in core, uncore, memroy controller, or
>
> memroy --> memory
Will change the typo.

> > +other componenets in a package. The feature is first implemented in Intel Sandy
>
> componenets --> components
>
Ditto.

> > +Bridge platform.
> > +
> Just for clarification - you mention a number of sensors, but unless
> I am missing something only the package sensor is implemented. Is that correct ?
Do you mean "only the core sensor"? I think the Sandy Bridge only implements
core sensor now. I would think in the future more sensors could be implemented.
If only core sensors are implemented, there is no need to have the package
level thermal hw MSRs because OS can get all package level thermal info by
enumerating all core sensors in a package.

>
> > +Temperature is measured in degrees Celsius and measurement resolution is
> > +1 degree C. Valid temperatures are from 0 to TjMax degrees C, because the actual
> > +value of temperature register is in fact a delta from TjMax.
> > +
> From the code, it seems that negative values can be reported. Is it guaranteed
> by the chip that (TjMax - MSR_IA32_TEMPERATURE_TARGET) >= 0 ? Otherwise, the
> minimum temperature would be (TjMax - 127).
>
> > +Temperature known as TjMax is the maximum junction temperature of package.
> > +Intel defines this temperature as 125C. At this temperature, protection
>
> Your driver bails out at TjMax >= 120, so there is some inconsistency.
> Also, it seems that this is not a constant, since you are reading
> MSR_IA32_TEMPERATURE_TARGET to get the value.
>
> Since the CPUs supporting the package sensor presumably also all support
> reading TjMax, maybe you can reword the above text to reflect this.
>
You are right. I forgot to update this part. I'll reword it.

> > +mechanism will perform actions to forcibly cool down the processor. Alarm
> > +may be raised, if the temperature grows enough (more than TjMax) to trigger
> > +the Out-Of-Spec bit. Following table summarizes the exported sysfs files:
> > +
> > +temp1_input - Package temperature (in millidegrees Celsius).
> > +temp1_crit - Maximum junction temperature (in millidegrees Celsius).
> > +temp1_crit_alarm - Set when Out-of-spec bit is set, never clears.
> > + Correct CPU operation is no longer guaranteed.
> > +temp1_label - Contains string "Pysical package id X", where X is physical
>
> Pysical --> Physical.
> I would suggest to drop the "id" for consistency. Core sensor names don't
> include "id" either.
>
> I am not sure if "physical" should be included in the first place.
> Also, above description suggests that future CPUs might add more sensors.
> If so, the name should probably reflect the location of the current sensor,
> ie be something like "Package core X" or "Core package X" or "Package X (core)".
How about just using "package X" to align with coretemp for now? Currently there
is no interface to tell which sensor it is.

Thanks.

-Fenghua


\
 
 \ /
  Last update: 2010-07-22 19:55    [W:1.805 / U:0.000 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site